Hydrol. Earth Syst. Sci., 22, 673–687, 2018https://doi.org/10.5194/hess-22-673-2018© Author(s) 2018. This work is distributed underthe Creative Commons Attribution 3.0 License.

Evaluation of uncertainties in mean and extreme precipitationunder climate change for northwestern Mediterranean watershedsfrom high-resolution Med and Euro-CORDEX ensemblesAntoine Colmet-Daage1,2,3, Emilia Sanchez-Gomez1, Sophie Ricci1, Cécile Llovel2, Valérie Borrell Estupina3,Pere Quintana-Seguí4, Maria Carmen Llasat5, and Eric Servat6

1CECI, CERFACS – CNRS TOULOUSE, Toulouse, France2WSP France, Toulouse, France3Hydrosciences Montpellier, Univ. Montpellier, Montpellier, France4Observatori de l’Ebre Fundació, Tarragona, Spain5Universitat de Barcelona, Barcelona, Spain6Institut Montpelliérain de l’Eau et de l’Environnement – IRD, Montpellier, France

Correspondence: Antoine Colmet-Daage (colmet@cerfacs.fr)

Received: 31 January 2017 – Discussion started: 14 March 2017Revised: 19 September 2017 – Accepted: 22 November 2017 – Published: 25 January 2018

Abstract. The climate change impact on mean and extremeprecipitation events in the northern Mediterranean regionis assessed using high-resolution EuroCORDEX and Med-CORDEX simulations. The focus is made on three regions,Lez and Aude located in France, and Muga located in north-eastern Spain, and eight pairs of global and regional climatemodels are analyzed with respect to the SAFRAN product.First the model skills are evaluated in terms of bias for theprecipitation annual cycle over historical period. Then futurechanges in extreme precipitation, under two emission sce-narios, are estimated through the computation of past/futurechange coefficients of quantile-ranked model precipitationoutputs. Over the 1981–2010 period, the cumulative pre-cipitation is overestimated for most models over the moun-tainous regions and underestimated over the coastal regionsin autumn and higher-order quantile. The ensemble meanand the spread for future period remain unchanged underRCP4.5 scenario and decrease under RCP8.5 scenario. Ex-treme precipitation events are intensified over the three catch-ments with a smaller ensemble spread under RCP8.5 reveal-ing more evident changes, especially in the later part of the21st century.

1 Introduction

The IPCC SRES report (IPCC, 2012) concludes with an in-crease in the frequency of heavy precipitation episodes overmost areas of the globe at the end of 21st century. In particu-lar, northwestern Mediterranean regions are often affected byextreme precipitation events that generate flash floods andcause serious damage (Ricard et al., 2012; Gaume et al.,2016). This climatically homogeneous region (Metzger et al.,2005) has been identified as a hotspot of climate change (CC)in the form of possible amplification of extreme precipitationassociated with a decrease in total precipitation (Gao et al.,2006; Giorgi, 2006; Giorgi and Lionello, 2008; Milano et al.,2013). Assessing the impacts of regional climate change onprecipitation is necessary in order to help and support policymakers develop strategies for future hydrological vulnerabil-ities like flash floods, but it constitutes a major challenge.

Global climate models (GCMs) are powerful tools to as-sess global-scale climate variability and change. GCMs haveallowed for better understanding of interactions betweenthe different components of the climate system (e.g., atmo-sphere, ocean, sea ice, continents). However GCMs gener-ally operate at coarse horizontal resolutions (100–250 km inthe atmospheric component); hence, they are not appropriatefor investigating the hydrological impacts of future extreme

Published by Copernicus Publications on behalf of the European Geosciences Union.

674 A. Colmet-Daage et al.: Evaluation of uncertainties

precipitation events at local scale, such as over hydrologicalwatersheds that are as small as hundreds of kilometers.

Interest in better representing climate variability andchange at local scales has driven the development of re-gional climate models (RCMs), which are currently able toperform dynamical downscaling of GCM at very high hori-zontal resolutions (∼ 10 km). RCMs run on limited area do-mains thereby allowing increased spatial resolution, and thusenabling better representation of surface heterogeneities andmesoscale atmospheric processes like convection (Fowleret al., 2007a). At European scale, collaborative researchprojects such as MERCURE (Hagemann et al., 2004), PRU-DENCE (Christensen et al., 2007), NARCCAP (Paulsen etal., 2009) and ENSEMBLES (van der Linden and Mitchell,2009) have contributed to further development and improve-ment of regional modeling. More recently, the internationalCORDEX (Coordinated Regional Climate Downscaling Ex-periment) initiative (Giorgi et al., 2009) has provided multi-model regional climate simulations at very high spatial reso-lution over different regions in the world. In particular, for thenorthwestern Mediterranean region, the EuroCORDEX andMedCORDEX sub-projects (EMCORDEX hereinafter) haveproduced present and future climate simulations at 12 km res-olution.

The merits of increased spatial resolution in RCMs havelargely been assessed in the literature. Comprehensive eval-uations of RCMs have been undertaken over the Euro-Mediterranean region by applying evaluation metrics tomean values of precipitation (Déqué and Somot, 2010; Fisheret al., 2012; Jacob et al., 2007; Kjellström et al., 2010; Kot-larski et al., 2005) as well as focusing on extreme precipi-tation associated with hydrological floods (Frei et al., 2006;Fowler et al., 2007b; Herrera et al., 2010; Kysel et al., 2012;Maraun et al., 2012). For recent EMCORDEX models, ini-tial evaluations over past periods have been conducted overEurope (Drobinski et al., 2016; Katragkou et al., 2015; Kot-larski et al., 2014). The latter evaluations mainly focusedon mean and extreme precipitation over the whole EM-CORDEX domain or in large regional boxes (e.g., France,the Alps, Mediterranean coastal regions, Morocco) usingsparse observation data sets. Prein et al. (2016) highlightedthat the added value of high-resolution models (12.5 km ver-sus 50 km) for the simulated mean and extreme precipitationdue to improved representation of orography and large-scaleconvection. Indeed, Drobinski et al. (2016) show the higherability of high-resolution RCMs to reproduce the Clausius–Clapeyron relation and thus precipitation-related processes.In the context of climate change, Jacob et al. (2013) showthat future climate projections performed by high-resolution(12.5 km) scenario under the Representative ConcentrationPathways 4.5 and 8.5 (RCP4.5 and RCP8.5, respectively)project higher daily precipitation intensities than GCMs, inparticular for RCP8.5. These results are consistent with theconclusions of Giorgi et al. (2016) over the Alps. Eventhough both GCM and RCM scenario experiments project

a reduction of summer precipitation over the Alps, increasedconvective rainfall due to enhanced potential instability re-lated to a finer representation of the orography over the Alpsis found in RCMs.

Together with increasing model resolution, high-resolution observation-based products have also beenrecently developed over different Mediterranean countries.Moving from CRU product at 50 km resolution (Harris etal., 2014) to E-OBS product at 25 km resolution (Haylock etal., 2008), reanalysis such as SAFRAN (Système d’AnalyseFournissant des Renseignements Atmosphériques à la Neige;Durand et al., 1993; Quintana-Seguí et al., 2008) or interpo-lated products such as SPAIN02 (Herrera et al., 2012, 2016)now provide precipitation products at a resolution compara-ble to those of the RCMs. In particular, the SAFRAN dataset available over France (Quintana-Seguí et al., 2008; Vidalet al., 2010) and Spain (Quintana-Seguí et al., 2016a, b)comprises a much larger observed data network than E-Obsor ERA-Interim (Dee et al., 2011), which were previouslyused for EMCORDEX model assessment (Cavicchia et al.,2016; Kotlarski et al., 2014). The SAFRAN-France data setwas used together with downscaled products issued fromCMIP5 (Coupled Model Intercomparison Phase 5) to assessfuture hydrological changes over France (Dayon, 2015;Quintana-Seguí et al., 2010, 2011). Themeßl et al. (2011)provide a review of downscaling methods. This implies sys-tematic bias correction of model precipitation before beingused as input for hydrological models, for instance in theframework of future flash flood simulation. Harader (2015)used the regional model ALADIN5.2 outputs at 12 kmresolution from CORDEX as well as SAFRAN-Franceproduct to describe future flash flood events over the Lezcatchment using a “futurization” method described in thefollowing.

The present study focuses on extreme precipitation overmesoscale northwestern Mediterranean watersheds withcomplex orography. Three watersheds of various sizes areinvestigated here: Lez and Aude located in southern France,and Muga, located in northeastern Spain. The goals of thisstudy are as follows:

– to assess RCM skills from the EMCORDEX multi-model ensemble in terms of mean and extreme precipi-tation values over past periods;

– to assess the influence of GCMs’ lateral boundaries con-dition on the RCMs’ skills;

– to evaluate future changes in precipitation extremes forfurther simulations of flash floods with an event-basedhydrological model over future periods.

For this purpose, the futurization approach proposed by Ha-rader (2015) is used. This method utilizes the computationof a past/future change coefficients of quantile-ranked RCMprecipitation outputs. Each multiplicative coefficient is then

Hydrol. Earth Syst. Sci., 22, 673–687, 2018 www.hydrol-earth-syst-sci.net/22/673/2018/

A. Colmet-Daage et al.: Evaluation of uncertainties 675

applied to each quantile-ranked short-term observed precip-itation event. It should be noted that the quantile-ranked ob-served precipitation is computed from SAFRAN daily pre-cipitation that have generated flash floods. The futurizationmethod is applied for each RCM in the EMCORDEX en-sembles, forced by two emission scenarios (RCP4.5 andRCP8.5). We thus explicitly take into account climate model-related uncertainty. In further work, the “futurized” precipi-tation events will be used with different hydrological mod-els, so that we explicitly take into account hydrology model-related uncertainty.

The paper is organized as follows, Sect. 2 includes abrief presentation of the EMCORDEX simulations, the ref-erence data sets and the statistical metrics applied to seasonalmean and extreme precipitation values. Section 3 presents theRCM evaluation in terms of mean and extreme precipitationvalues over the present period when the global scale is pre-scribed by ERA-Interim. Section 4 analyzes present climatesimulations to understand the role of the GCMs in drivingthe RCMs. The impacts of climate change on precipitationare then examined in Sect. 5. Conclusions and perspectivesare finally given in Sect. 6.

2 Data and methods

2.1 The river catchments

In the current study, the futurization approach is appliedover three Mediterranean watersheds with different charac-teristics and external influences. The Lez, Aude and Mugacatchments displayed in Fig. 1 are frequently subject to flashfloods that cause considerable damage to surrounding areasand cities.

The upstream part of the Lez watershed, shown in redin Fig. 1, is located 15 km north of the city of Montpellierand covers 114 km2. The landscape is dominated by garriguevegetation, very common in the Mediterranean countries.The spring of the Lez River is the resurgence of a karsticaquifer of about 380 km2. The karst aquifer plays an impor-tant role in water resources in the basin, and the karst out-crops actively participate in flash flood dynamics (Raynaudet al., 2016). Cumulative annual rainfall is around 909 mm,which, on average, falls 60 days per year (Coustau, 2011;Harader, 2015). The Lez catchment is frequently subject toflash floods caused by extreme precipitation episodes region-ally known as the “Cévenols” events (Ducrocq et al., 2008;Nuissier et al., 2008, 2011). Cumulative extreme precipita-tion can locally reach 600 mm in 24 h within the river catch-ment (Boudevillain et al., 2011).

The Aude watershed, shown in brown in Fig. 1, coversmore than 5000 km2 upstream the nearby city of Narbonne.The Aude River springs at the Pyrenees and flows along thecatchment for 223 km before entering the Mediterranean Sea.The Orbieu and the Fresquel are its major tributaries. The

Aude catchment is surrounded by several mountain chainssuch as the Cevennes massif to the north and the Pyre-nees to the south. The catchment is mainly dominated bya Mediterranean climate, but large climate contrasts can befound over its sub-watersheds. A severe flash flood episodeoccurred in November 1999 (more than 200 mm of rain in24 h over a major part of the catchment) and caused severedamage over an extended region (Aude, Tarn, Pyrénées Ori-entales, Hérault; Estupina, 2004; Bechtold and Bazille, 2001;Ducrocq et al., 2003; Gaume et al., 2004).

Finally, the Muga catchment (shown in purple in Fig. 1) lo-cated in northeastern Spain over Catalonia covers 854 km2.The Muga River is about 58 km long, between the Pyre-nees (maximum altitude of the catchment, 1214 m) and theGulf of Roses. This catchment is usually affected by heavyprecipitation associated with convective events, with annualprecipitation average between 700 mm in the upper part and530 mm at the mouth, and daily precipitation of 200 mm forreturn period of 10 years (Llasat et al., 2009, 2014). Thepreviously mentioned November 1999 flash flood event ledto a historical peak discharge recorded near the mouth ofthe Muga River, with 925 m3 s−1 compared to an averagevalue of 3.34 m3 s−1. The Muga catchment was affected by26 severe flood events between 1982 and 2010 (Llasat et al.,2014).

It should be noted that, in this paper, statistical analysis iscarried out over the Lez, Aude and Muga catchment areas aswell as over more extended regional boxes also displayed inFig. 1.

2.2 RCM Euro-Mediterranean CORDEX simulations

The set of EMCORDEX simulations used in this study aresummarized in Table 1. The five RCMs used are presentedwith the main reference papers (in particular with respect toboundary layer and convection schemes). The EMCORDEXcommunity has provided three types of RCM simulationswith driving conditions (also detailed in Table 1) issued fromeither ERA-Interim or GCMs simulations over past and fu-ture periods:

– The evaluation simulations (EVAL hereinafter). The lat-eral boundary conditions (LBCs) are driven by ERA-Interim reanalysis (Dee et al., 2011) over 1981–2010.These simulations are used to evaluate the RCM intrin-sic biases.

– The historical simulations (HIST hereinafter). TheLBCs are issued from numerical experiments per-formed with four different GCMs and extracted fromthe CMIP5 historical archive. These historical simula-tions represent the climate conditions over 1976–2005.

– The future climate scenario simulations (RCP here-inafter). In this case, the LBCs correspond to four fu-ture climate scenarios (from four different GCMs from

www.hydrol-earth-syst-sci.net/22/673/2018/ Hydrol. Earth Syst. Sci., 22, 673–687, 2018

676 A. Colmet-Daage et al.: Evaluation of uncertainties

Figure 1. Mean relative seasonal precipitation bias (%) in all the RCMs forced with ERA-Interim for the period 1981–2010. The top left panelshows the horizontal pattern of seasonal precipitation provided by the SAFRAN reference data set (mm season−1). Only the September–October–November (SON) season is shown here. On the SAFRAN map (a), the colored rectangles represent the regional boxes, while theoutline represents the hydrological catchment. The Cevennes box and the Lez catchment are in red, the Aude box and catchment in brown,and the Muga box and catchment in purple.

Table 1. The parameterization and physics scheme related to the RCM (0.11◦) used in the present study are presented in the five first lines.The lateral boundary conditions (LBCs) for each type of simulations are presented with four different GCMs of CMIP5. The last row liststhe name used for each GCM–RCM pair throughout this study.

RCM ALADIN5.2 ALADIN5.3 RCA4 HIRHAM5 RAMO22E

Institution Météo-France Météo-France SMHI DMI KNMI

Boundarylayer scheme

Ricard andRoyer (1993)

Ricard andRoyer (1993)

Cuxart et al. (2000) Louis (1979) Lenderinkand Holtslag(2004);Siebesma etal. (2007)

Convection Mass fluxBougeault(1985)

Mass fluxBougeault (1985)

Kain and Fritsch (1990, 1993);Kain (2004);Jones and Sanchez (2002)

Tiedtke (1989) Tiedtke (1989);Nordeng (1994);Neggers et al.(2009)

Mainreferences

Colin et al.(2010);Hermann et al.(2011)

Colin et al. (2010);Hermann et al. (2011)

Samuelsson et al. (2011);Kupiainen et al. (2011)

Christensenet al. (2008)

Meijgaardvan et al. (2012)

Evaluation ERA-INTERIM (reanalysis)

Lat

eral

boun

dary

cond

ition

s simulation

Historicalsimulation

RCP4.5simulation

CNRM-CERFACS-CNRM-CM5 MOHC-Had

MPI-M-MPI-

ICHEC-EC-EARTH

RCP8.5simulation

GEM2-ES

ESM-LR

GCMmembers

r8i1p1 r1i1p1 r1i1p1 r1i1p1 r1i1p1 r12i1p1 r3i1p1 r1i1p1

Name of pairGCM_ RCM

CNRM-CM5_ ALADIN52

CNRM-CM5_ ALADIN53

CNRM-CM5_ RCA4

MOHC_RCA4

MPI_RCA4

ICHEC_RCA4

ICHEC_HIRHAM5

ICHEC_RACMO22E

Hydrol. Earth Syst. Sci., 22, 673–687, 2018 www.hydrol-earth-syst-sci.net/22/673/2018/

A. Colmet-Daage et al.: Evaluation of uncertainties 677

CMIP5) under RCP4.5 and RCP8.5 (Clarke et al., 2007;Riahi et al., 2007; Meinshausen et al., 2011; van Vuurenet al., 2011) over 2011–2040, 2041–2070 and 2071–2100. These simulations cover 30-year time slice peri-ods like the HIST simulation. Thus they can be statisti-cally compared to assess future changes.

This results in eight pairs of RCM–GCM simulations ana-lyzed in this study, for HIST and RCP as indicated in the lastrow of Table 1. Only simulations at 12 km for which EVAL,HIST and both RCP experiments were available at the begin-ning of the study were considered.

2.3 The reference data set: SAFRAN

SAFRAN reanalysis provides daily precipitation data for theperiod 1958–2008 over France and Spain on an 8 and 5 kmgrid, respectively (Vidal et al., 2010, Quintana-Seguí et al.,2016a, b). SAFRAN-France was built by using data from3675 selected rain gauges that were gridded through an op-timal interpolation algorithm described in Quintana-Seguí etal. (2008). The great number of rain gauges considered inSAFRAN and its high spatial resolution produce more accu-rate precipitation analyses over France and Spain than thoseproposed by other products commonly used for model as-sessments, such as CRU (Harris et al., 2014) and E-OBS(Haylock et al., 2008). A recent study shows that, for precip-itation, the performance of SAFRAN is very similar to thatof Spain02 (Quintana-Seguí et al., 2016b). SAFRAN dataset has been evaluated using the Météo-France and Span-ish State Meteorological Agency (AEMET) gauging stationnetwork as well as independent data (Quintana-Seguí et al.,2008, 2016a, b; Vidal et al., 2010).

SAFRAN product is used here as a reference data set toevaluate the simulated precipitation from EVAL and HISTensembles. For that purpose, the 12 km RCM outputs andalso SAFRAN Spain were regridded on the 8 km SAFRAN-France grid using the ESMF (Earth System Modeling Frame-work) bilinear regridding method. The impact of interpola-tion has been largely investigated in the literature, for in-stance in Diaconescu et al. (2015), and it should be quanti-fied before interpretation of interpolated fields. In the presentstudy, the impact of a bilinear scheme for the interpolationof precipitation fields from 12 km simulation grid to 8 kmSAFRAN grid was assessed from an analytical test casebased on the cosine function of longitude and latitude. Thecomparison is achieved by mapping the analytical field ontothe 12 km fields and interpolated it onto an 8 km grid withthe analytical field mapped onto the 8 km. The quadratic er-ror reaches 5.10−5; the impact of interpolation was thus ne-glected in the following.

2.4 Statistical metrics to evaluate RCM performance

Given the small size of the Lez, Aude and Muga catchments(3, 84 and 11 SAFRAN grid points, respectively), and in or-

der to allow for proper statistical analysis, the evaluation ofmean precipitation (seasonal and annual cycle) was achievedon larger regional boxes (excluding sea grid points) shownin Fig. 1. These regional boxes have been selected accord-ing to regions of homogeneous climate conditions. In theparticular case of the Lez catchment, the RCM precipitationis evaluated over the entire Cevennes region. However forall three catchments, extreme precipitation metrics are com-puted over grid points that are strictly inside the catchments.The precipitation extremes are analyzed with respect to the90th to 99.9th quantiles of the daily precipitation distributiondiscretized as follows:

– one point per 1 quantile rank from 90 to 95th;

– one point per 0.5 quantile rank from 95 to 98th;

– one point per 0.2 quantile rank from 98 to 99th;

– one point per 0.1 quantile rank from 99 to 99.9th.

Short-term observed precipitation events are thereby com-pletely covered and evenly distributed with this quantile dis-cretization (not shown). Quantiles are computed consideringall the days (rainy days and dry days), thus allowing for acomparison of precipitation quantiles between the differentRCM–GCM pairs (Giorgi et al., 2016; Schär et al., 2016).However, it should be noted that when precipitation is below0.1 mm day−1, the precipitation is set to 0 so that the cumu-lative precipitation is not affected by overrepresentation oflight rainy days in the models (Harader et al., 2015; Tram-blay et al., 2013, Paxian et al., 2015).

In addition to the classical metrics (spatial bias, annualcycle bias, quantile–quantile plot), two original metrics areused in this study. First, assuming additivity between GCMand RCM errors, the impact of the GCM bias on the RCMsolution can be diagnosed by computing the difference 1B

between the HIST and the EVAL precipitation bias with re-spect to SAFRAN:

1B =HIST−SAFRAN

SAFRAN−

EVAL−SAFRANSAFRAN

. (1)

The 1B criteria corresponds to the bias in the annual cycle ofprecipitation simulated with the RCMs that is strictly relatedto the influence of the lateral boundary condition imposed bythe GCM. A high positive value indicates overestimation ofthe total monthly precipitation, and a negative value indicatesunderestimation of the total monthly precipitation.

Secondly, in the present study, change coefficients be-tween the past (HIST) and future precipitation (RCP) quan-tile distributions are computed. For that purpose, precipita-tion from HIST and RCPs is quantile-ranked, and for eachquantile rank “qi”, a change coefficient “Aqi” for precipita-tion intensities between HIST and RCP is computed as fol-lows:

Aqi=Pqi(RCP)

Pqi(HIST), (2)

www.hydrol-earth-syst-sci.net/22/673/2018/ Hydrol. Earth Syst. Sci., 22, 673–687, 2018

678 A. Colmet-Daage et al.: Evaluation of uncertainties

where Pqi(RCP) and Pqi(HIST) represent the values of thequantiles for HIST and RCP, respectively.

All metrics are computed on a seasonal basis, consider-ing hereinafter four seasons: autumn (September, Octoberand November), winter (December, January and February),spring (March, April and May) and summer (June, July andAugust). Only autumn and spring results are presented in thefollowing section.

3 Analysis of RCM evaluation simulations

In this section the RCM precipitation biases are diagnosedthrough the comparison between the EVAL simulations andSAFRAN. The spatial pattern for the mean cumulative pre-cipitation, the annual cycle and the extreme values are inves-tigated.

3.1 Spatial bias pattern

Figure 1 shows the spatial distribution of the cumulative pre-cipitation normalized difference between each RCM fromthe EVAL ensemble and SAFRAN, averaged over the previ-ous 30 years in autumn. The top-left panel in Fig. 1a displaysthe mean cumulative precipitation for SAFRAN-France andSAFRAN-Spain reference data sets. Large values of thecumulative precipitation (greater than 400 mm season−1)

are observed over mountainous regions, in particular theCevennes and the Pyrenees chains. Lower cumulative pre-cipitation values are observed in the valleys (Garonne, Aude)and over the coastal regions.

In general, the cumulative precipitation is overestimatedfor most RCMs over the mountainous regions (+30 %)and underestimated over the Mediterranean coastal region(−30 %), as shown in Fig. 1b to f, most likely because ofimperfect representation of the orography as well as theparameterization of the convection scheme. Note that theRACMO22E pattern of precipitation bias differs from theother RCM patterns. Indeed, slight overestimation is ob-served over the south of the Pyrenees and the Cevennes(+20 %), and almost no bias is observed over the valleys.

We focus now on the precipitation bias over river catch-ments that are analyzed in the present study: for Cevennes,the cumulative precipitation is overestimated by 20 % in thesouthwest mountainous region and underestimated by 30 %in the northeast valley region for all RCMs (Fig. 1b, c, e, f)except for RACMO22E (Fig. 1d).

For the Aude region, no relevant bias is presented by AL-ADIN52 (Fig. 1b), ALADIN53 (Fig. 1c) and RACMO22E(Fig. 1d). A positive bias (+40 %) is presented by RCA4(Fig. 1e) in the western region under continental influence,while no relevant bias is presented in the eastern region un-der Mediterranean influence. HIRHAM5 (Fig. 1f) displaysa strong positive bias (+50 %) in the high-elevation areas

Figure 2. (a) Annual cycle of precipitation over Aude box withSAFRAN data set (b). Bias of the annual cycle of precipitation sim-ulated by the five RCMs with respect to SAFRAN’s annual cycle ofprecipitation for the period 1981–2010 (no units), over the Audebox. The colored lines with different markers represent the biasesof RCM-simulated precipitation with respect to SAFRAN data set(black line).

(Pyrenees in the southwest, and Black mountain in the north-west) and a strong negative bias elsewhere.

The mean cumulative precipitation is underestimated(−30 %) by all RCMs over the Muga region, except forRACMO22E.

For other seasons, the mean cumulative precipitation pat-tern tends to be overestimated over all three regions in spring(not shown), and a strong positive bias (+50 %) is presentedby ALADIN52 and ALADIN53 over the Pyrenees in sum-mer (not shown).

3.2 Annual cycle of precipitation bias

The 30-year climatology for the monthly cumulative precipi-tation is spatially averaged over each region of interest andnormalized by the SAFRAN monthly climatology. Figure2a displays the annual cycle of precipitation over the Audebox for SAFRAN data set. Figure 2b displays the bias ofthe annual cycle over the Aude region for each RCM. Glob-ally speaking, the RCM bias and the inter-model spread aresmaller in autumn than in summer. This can be explained bythe influence of the large-scale atmospheric circulation be-ing weaker in summer, the control exerted by the LBCs onthe RCM is reduced in this season and the RCM solutionhas more degrees of freedom to deviate from the large-scaleforcing (Déqué et al., 2012; Lucas-Picher et al., 2008).

Hydrol. Earth Syst. Sci., 22, 673–687, 2018 www.hydrol-earth-syst-sci.net/22/673/2018/

A. Colmet-Daage et al.: Evaluation of uncertainties 679

ALADIN52 and ALADIN53 clearly overestimate the cu-mulative monthly precipitation in late spring and summer.This is likely due to a large presence of low precipitation daysin the RCM precipitation distribution (0 to 10 mm day−1)

compared to SAFRAN distribution (not shown). This behav-ior is also observed over the Cevennes and Muga regions(not shown). Consistent with Fig. 1d, RACMO22E displaysa low bias in the annual cycle. Similar behavior is observedfor HIRHAM5, suggesting an error compensation betweenpositive and negative spatial biases displayed in Fig. 1f. Forthe Muga and the Cevennes boxes (not shown), HIRHAM5presents a negative bias. Finally, RCA4 simulates a large pos-itive bias (+30 %) for the entire annual cycle that tends tointensify over the Alps and the western Pyrenees.

3.3 Extreme precipitation bias

The ability of the EVAL simulations to represent the extremeof precipitation is investigated through a quantile–quantileanalysis with respect to SAFRAN. This analysis is performedfor each grid point within the Lez, Aude and Muga catch-ments. The highest-order quantiles (95 to 99th) are displayedin Fig. 3 for each RCM and SAFRAN in autumn.

For the Lez, Aude and Muga catchments, the RCMs under-estimate the higher-order quantiles. This underestimation ismore important for the Muga catchment, where the extremeprecipitation intensities are stronger.

ALADIN52 underestimates the upper precipitation quan-tiles; for instance, the intensity of the 99th quantilesonly reaches 70 mm day−1 compared to 140 mm day−1

for SAFRAN. ALADIN53 performance is slightly im-proved with respect to ALADIN52, especially for the Audeand Muga catchments. RACMO22E and RCA4 underesti-mate the extreme precipitation above the 99.5th quantiles.HIRHAM5 provides a satisfactory description of all extremequantiles for the three catchments.

3.4 Summary of means and extremes analysis

The mean and extreme precipitation is investigated over theAude, Lez and Muga catchments. Largest biases in mean pre-cipitation simulated by the regional climate models are lo-cated over the mountainous (positive bias) and coastal (nega-tive bias) regions. These biases are stronger during the sum-mer season, when the control exerted by the LBCs on theRCM is weaker, due to a reduction of the large-scale circu-lation and North Atlantic inflow. Overall, while the climatevariability is covered by the spread of EMCORDEX ensem-ble of RCMs, each RCM simulates plausible precipitation.

These results are coherent with specific studies for meanand extreme precipitation as cited hereafter. ALADIN52 pre-cipitation biases are consistent with those obtained by Ha-rader (2015), and slightly reduced in ALADIN53. The RCA4overestimation of mean precipitation is in accordance withthe results showed in Prein et al. (2016). The underestima-

tion of extreme precipitation diagnosed in our study is pos-sibly even more severe than our reference, SAFRAN, foundto underestimate observed extreme precipitation (Quintana-Seguí et al., 2008a). This last point was confirmed througha comparison between pluviometer data and SAFRAN overour catchments of interest.

Parameterization and physical processes that are involvedin the generation of low precipitation differ from those in-volved in stronger precipitation events. Thus the precipita-tion bias varies according to the precipitation intensity. Quan-tile classification is thereby relevant for the determination ofpast/future precipitation change.

4 Analysis of RCM historical simulations

RCM intrinsic biases, determined from the EVAL ensemble,have been characterized and described in the previous sec-tion. Here we analyze the historical simulations (HIST), per-formed with the GCM–RCM pairs, as described in Sect. 2.2.In this case, GCM biases are expected to affect the RCM sim-ulation. A difficult challenge is to understand how GCM andRCM biases are interacting (Dequé et al., 2012). In this studywe make the assumption of additivity between the impact ofthe GCM biases and the RCM intrinsic biases. In this study,four different GCM forcings are considered: CNRM-CM5,ICHEC, MOHC and MPI (see Table 1).

4.1 Annual cycle of precipitation bias

Figure 4 displays the annual cycle of 1B over the Aude wa-tershed for each GCM–RCM pairs. The color code refersto the GCM (for instance for CNRM-CM5 forcing), whilethe markers refer to the RCM (for instance star for AL-ADIN53). From Fig. 4 it is evident that CNRM-CM5 forcingleads to systematic overestimation of summer precipitation,hence proving that the positive precipitation bias identifiedin EVAL (Sect. 3, Fig. 2 for ALADIN52, ALADIN53 andRCA4) is enhanced in HIST simulations.

ICHEC large-scale conditions induce no significantchanges on RACMO22E and RCA4 errors, while they leadto strong overestimation of the HIRHAM5 precipitation. Ex-cept when forced by CNRM-CM5, the bias of GCM–RCA4pairs is similar to the intrinsic bias of RCA4, as displayed forall three, MPI, MOHC and HIRHAM5. Illustration is pro-vided for the Aude box in Fig. 4, but similar results are ob-tained for the Cevennes and the Muga boxes.

It should be noted that previous results might also be ex-plained by bias compensation between GCM impacts andRCM intrinsic bias. Despite the GCM–RCM deficienciesshown here, all the GCM–RCM pairs display plausible pre-cipitation and are then considered in the following for futurechange analysis over mean precipitation annual cycle.

www.hydrol-earth-syst-sci.net/22/673/2018/ Hydrol. Earth Syst. Sci., 22, 673–687, 2018

680 A. Colmet-Daage et al.: Evaluation of uncertainties

Figure 3. Quantile–quantile diagram of daily precipitation in the cells of the three catchments for the period 1981–2010. EVAL-simulatedprecipitation (colored lines with different markers) is compared to SAFRAN (black line). The x axis represents the precipitation quantilevalues with respect to the SAFRAN reference data. The y axis represents the precipitation intensity simulated by the EMCORDEX modelsfor the same quantiles. If a colored line is above/below the black line, the corresponding RCM over/underestimates quantile intensities withrespect to SAFRAN. Units are in mm day−1. The colored dots represent, from left to right, the 90, 95, 97, 99, 99.5 and 99.9th quantiles.

Figure 4. Bias on the annual cycle RCM-simulated precipitationinduced by GCMs lateral boundary conditions. Aude regional boxis shown here for the period 1976–2005. This bias is estimatedthrough the computation of the difference between HIST and EVALsimulated precipitation with respect to SAFRAN. 1B (no units)represent this bias with 1B = HIST−SAFRAN

SAFRAN −EVAL−SAFRAN

SAFRAN .Colored lines refer to the GCMs, and the markers refer to the RCMsdriven. If a colored line is above (below) the black line, the corre-sponding GCM induce overestimation (underestimation) of the pre-cipitation simulated by the RCM.

4.2 Extreme precipitation bias

Figure 5 is the counterpart of Fig. 3; it displays quantile–quantile diagrams of precipitation in autumn for the HISTensemble over the Lez, Aude and Muga catchments. First,the spread amongst the HIST simulations displayed in Fig. 5is larger, except for the Muga catchment, larger than the onegenerated by the EVAL ensemble (Fig. 3) that the EVAL ones(Fig. 3), and extreme quantiles tend to be systematically un-derestimated over all three catchments of interest.

Generally, CNRM-CM5 forcing leads to underestimationof SAFRAN quantiles, thus enhancing the RCMs’ intrinsicbiases displayed in Fig. 3. MPI and ICHEC provide good

quantile distribution for all RCMs except beyond 99.5thquantile. In more detail, ICHEC-RCA4 extreme quantilesare slightly overestimated over the Aude catchment, and ingood agreement with SAFRAN over the Muga catchment.MOHC-RCA4 provides good statistics over Aude and Mugacatchments, while it overestimates all quantiles over the Lezcatchment.

To conclude, the impact of the GCM on RCMs tends tosystematically intensify the underestimation of extreme pre-cipitation values. However, the HIST ensemble remains con-sistent with SAFRAN statistics and thus will be used in thefollowing step, in order to estimate future changes over ex-treme precipitation over the different catchments.

5 Effect of climate change on precipitation

5.1 Annual cycle of precipitation

Figure 6 presents the precipitation annual cycle simulatedfor HIST computed over 1976–2005 period and the scenarioRCP4.5 and RCP8.5 computed over 2071–2100 period, forthe Cevennes, Aude and Muga boxes.

The results reveal stronger mean precipitation change be-tween RCP4.5 and RCP8.5 than for the 2011–2040 and2041–2070 periods. The annual cycle of the RCP4.5 ensem-ble is similar to HIST one in terms of mean and spread,suggesting that radiative forcing in RCP45 seems to have aweak impact on monthly averaged precipitation. In contrast,the annual cycle of the RCP8.5 ensemble displays a gen-eral decrease in mean precipitation from April to Octoberfor the three river catchments. These results are consistentwith the conclusions from the study of Jacob et al. 2013 overEMCORDEX domain. As previously mentioned the spreadof the ensemble is larger in summer when the LBCs exerta weaker control in the RCM domain. It should be noted

Hydrol. Earth Syst. Sci., 22, 673–687, 2018 www.hydrol-earth-syst-sci.net/22/673/2018/

A. Colmet-Daage et al.: Evaluation of uncertainties 681

Figure 5. Quantile–quantile diagram of daily precipitation in the cells of the three catchments for the period 1981–2010. HIST-simulatedprecipitation (colored lines) is compared to SAFRAN (black line). The x axis represents the precipitation quantile values with respect to theSAFRAN reference data. The y axis represents the precipitation intensity simulated by the EMCORDEX models for the same quantiles. If acolored line is above/below the black line, the corresponding RCM over/underestimates quantile intensities with respect to SAFRAN. Unitsare in mm day−1. The colored dots represent, from left to right, the 90th, 95th, 97th, 99th, 99.5th and 99.9th quantiles.

that these results hold for the three boxes of interest that areof significantly different sizes. This suggests that the high-resolution EMCORDEX simulations can be confidently usedto investigate precipitation at local scale.

As mentioned in the previous section, the analysis of theannual cycle of HIST simulations emphasized that the asso-ciation of GCM and RCM models induces a different pre-cipitation bias that the intrinsic RCM biases (comparison be-tween black and blue curve in Fig. 6). Nevertheless, the pre-cipitation field issued from HIST is not bias corrected here.They are used, together with the RCP simulations, to esti-mate the change coefficient between past and future valuesof the quantile of the precipitation distribution.

5.2 Change coefficients for extreme precipitation

The change coefficients between quantile-ranked precipita-tion presented in Eq. (2) are displayed in Fig. 7. They allowthe estimation of the changes between the past (HIST) andfuture precipitation (RCP) quantile distributions

Following Eq. (2), a change/transfer coefficient greaterthan 1 for a given quantile indicates an increase of the fu-ture precipitation value associated with this quantile. On theother hand, Aqi < 1 means a decrease in RCP precipitationwith respect to HIST.

The Aqi coefficients are computed for each GCM–RCMpair over the Lez, Aude and Muga catchments. The multi-model approach adopted here allows for estimating an uncer-tainty in the values of the change coefficients. The Aqi coef-ficients are represented in Fig. 7 over the upper level quantilerange (90–99.9) for RCP4.5 and RCP8.5 in autumn for thethree river catchments and for 2041–2070 and 2071–2100future time periods. The ensemble mean for Aqi computedamongst the GCM–RCM pairs is compared to the associatedensemble spread (standard deviation) in Fig. 7.

The interesting result here is that while RCP simula-tions tend to decrease mean precipitation with respect toHIST simulation (Fig. 6), extreme precipitation events areintensified for both time periods over the three catchments.Globally speaking, the mean change coefficient for RCP4.5and RCP8.5 is similar except for the Lez (2041–2070)and the Muga (2071–2100) catchments, where RCP8.5 dis-plays biggest changes. The ensemble spread from RCP8.5 issmaller than the one from RCP4.5, meaning that the changein extreme precipitation displays higher level of certainty un-der RCP8.5 scenario, which can be explained by stronger ra-diative forcing in RCP85 compared to RCP45. Few modelsindicate a decrease in extreme precipitation for RCP4.5 over2041–2070 period. In contrast, for 2071–2100, all the GCM–RCM members agree on an increase in extreme precipita-tion for both RCP4.5 and RCP8.5 scenarios. For instance,the mean change coefficient over 2071–2100 reaches 1.15 forthe 99.5th quantile over the Aude catchment for both RCPs,while RCP4.5 spread is about 0.3, versus 0.1 for RCP8.5. Interms of precipitation, a change coefficient of 1.35 over theLez catchment for the 99.9th quantile of RCP8.5 representsan increase from 140 to 189 mm day−1.

6 Discussion

The present study assessed the intensification of extreme pre-cipitation events under climate change on small Mediter-ranean catchments using high-resolution RCM simulations(∼ 12 km) from the EMCORDEX exercise. It was shown thatover the past period (1976–2005), EVAL simulations (RCMdriven by ERAI) and HIST simulations (RCM is drivenby a GCM) underestimate extreme events with respect toSAFRAN data set.

Despite SAFRAN being the best gridded observation dataset available covering the region of interest as explained in

www.hydrol-earth-syst-sci.net/22/673/2018/ Hydrol. Earth Syst. Sci., 22, 673–687, 2018

682 A. Colmet-Daage et al.: Evaluation of uncertainties

Figure 6. Annual cycle of mean precipitation (mm month−1)simulated in the three regional boxes (Cevennes, Aude, Muga;see Fig. 1). Solid lines represent the ensemble means computedamongst the eight GCM–RCM pairs. The shaded areas representthe ensemble spreads characterized by the standard deviation. HISTensemble mean (1976–2005) is plotted in blue. RCP4.5 and RCP8.5ensemble means (2071–2100) are respectively plotted in green andred. SAFRAN annual cycle computed over 1981–2010 is plotted inblack.

Sect. 2.3, the underestimation highlighted here is potentiallymore severe as SAFRAN is itself deemed to underestimateobserved extreme precipitation events (Quintana-Seguí et al.,2008a). Indeed, complementary analysis (not shown here) onthe Aude and Lez catchments highlighted that SAFRAN un-derestimates precipitation quantiles beyond the 95th quan-tile by up to 30 % with respect to pluviometer data from theMétéo-France network. It should be noted that comparisonbetween the SAFRAN gridded product and the pluviome-ter sparse product is challenging and that these products arenot independent. However, more representative computation

of observed quantiles from pluviometer data set in place ofSAFRAN ones is one potential gain from the present study.This paves the way towards a time-varying quantile classi-fication with an intra-daily cycle, for instance by using 3 hpluviometer data. Such an intra-daily cycle could also be rep-resented in the past/future change coefficients using 3 h EM-CORDEX precipitation outputs. Working on higher tempo-ral resolution data (observation and simulation) would leadto a futurization method more consistent with flash floodtimescales than the daily current method.

Future changes in precipitation are assessed through amulti-model approach focusing on the mean and standard de-viations of the EMCORDEX ensemble. In order to reducethe model uncertainty, different classifications and cluster-ing methods, based on advanced ensemble statistics, havebeen tested to exclude the outlier models from a specific en-semble, or to apply weightings procedures in the ensemblemean computation (Boberg and Christensen, 2012). How-ever, as discussed in Reifen and Toumi (2009) and Knutti etal. (2010), model performances in the past do not necessarilyrelate to model performances in the future and the choice ofthe criteria for multi-model studies should be further investi-gated.

As previously explained in Fig. 6, the monthly averagedprecipitation simulated by the RCP8.5 ensemble decreaseswith respect to the HIST monthly averaged precipitation overthe past period, especially from April to October for the Lez,Aude and Muga catchments. For the Aude and Muga boxes,there is a shift in the annual cycle; the peak occurs earlierin spring (in April instead of May). This suggests that achange in precipitation amplitude and in temporal season-ality should be expected. Llasat and Puigcerver (1997) ex-plained that intense precipitation events are mostly due toconvective rainfall in autumn and to global circulation inspring. The maximum number of convective days and ra-tio between convective and total precipitation is recorded be-tween August and September in la Boadella, and a positivetrend (5 % error level) in the annual number of convectivedays has been founded in the Muga catchment (Llasat et al.,2016). The change in spring precipitation is thus most likelydue to changes in global fluxes that have different impacts onwestern coastal regions (Aude and Muga) and on southerncoastal regions (Cevennes). Indeed, Nuissier et al. (2011) ex-plained that strong precipitation events over the west coast ofthe Mediterranean are mostly correlated with easterly fluxes,while strong precipitation events in the northern coastal re-gion are correlated with south to southeasterly fluxes. Cassouet al. (2016) reported that climate change will have a greaterimpact on easterly fluxes; this could explain the change in theannual cycle observed in Aude and Muga regions only. Thishypothesis should be validated with an analysis of changesin geopotential fields.

Hydrol. Earth Syst. Sci., 22, 673–687, 2018 www.hydrol-earth-syst-sci.net/22/673/2018/

A. Colmet-Daage et al.: Evaluation of uncertainties 683

Figure 7. Change coefficients (Aqi) over the 90–99.9 quantile range computed for each GCM–RCM pairs over the Lez, Aude and Mugacatchments during the autumn season (SON). The no change line (ai= 1) is also displayed in solid black. The thick solid lines represent theensemble mean for (Aqi), and the associated ensemble spread is presented by the shaded areas (standard deviation). RCP4.5 and RCP8.5 arerespectively plotted in blue and red. For clarity purposes, the scale of the x axis was distorted according to the quantile discretization.

7 Conclusions

A futurization approach is presented in this article; it con-sists of the computation of a past/future change coefficientsapplied to quantile-ranked RCM precipitation outputs. Weuse the EMCORDEX ensemble to estimate the model andscenario uncertainty. The study focuses over the Lez, Audeand Muga Mediterranean catchments. As a first step, EM-CORDEX models’ skills are evaluated in terms of mean andextreme precipitation over the present climate period 1976–2005. This analysis legitimizes the use of the EMCORDEXmodels for past/future changes assessment.

It has been shown that cumulative precipitation is over-estimated over the mountainous regions and underestimatedover the coastal regions in autumn. Extreme events beyond95th quantiles are underestimated. GCM forcing tends to en-hance RCM underestimation of extreme precipitation eventsover the three catchments. Climate change impact is inves-tigated from the RCP4.5 and RCP8.5 EMCORDEX simu-lations. In comparison with the present period, the monthlyaveraged precipitation decreases in spring and summer forRCP8.5. Past/future change coefficients computed from theEMCORDEX ensemble display an increase in extreme pre-

cipitation event intensity (beyond 90th quantile). This resultis stronger over the end of 21st century, for RCP8.5, for allcatchments of interest: all models within the ensemble agreeon change coefficients larger than 1. The multi-model ap-proach developed here allows for quantifying the uncertaintyrelated to the past/future change coefficients. As a major con-clusion of this study, we have shown with a high degree ofconfidence that all RCM models in EMCORDEX ensembleforecast an increase in extreme precipitation events.

In the future, change coefficients could be used to providea kind of futurized extreme precipitation event that occurredin the past. In further studies, the hydrological impact ofthis futurized precipitation will be assessed using rainfall–runoff models over the Lez, Aude and Muga catchments.This generic method could also be applied to other catch-ments.

Data availability. The raw data corresponding to the precipitationsimulated by the 8 RCMs from EMCORDEX can be accessed onthe CORDEX or the ESGF websites. The postreated data are con-served by the CECI-CERFACS and WSP France. To consult it,please contact the authors.

www.hydrol-earth-syst-sci.net/22/673/2018/ Hydrol. Earth Syst. Sci., 22, 673–687, 2018

684 A. Colmet-Daage et al.: Evaluation of uncertainties

Competing interests. The authors declare that they have no conflictof interest.

Acknowledgements. This work is an initiative of WSP-France. Thefinancial support for this work has been provided by WSP-Franceunder the CIFRE contract 2015/005. This paper was writtenunder the framework of the International HYMEX project and theSpanish HOPE (CGL2014-52571-R) project. We acknowledge theWorld Climate Research Programme’s Working Group on RegionalClimate, and the Working Group on Coupled Modelling, formercoordinating body of CORDEX and responsible panel for CMIP5.We also thank the climate modeling groups (listed in Table 1 of thispaper) for producing and making available their model output. Wealso acknowledge the Earth System Grid Federation infrastructurean international effort led by the U.S. Department of Energy’sProgram for Climate Model Diagnosis and Intercomparison,the European Network for Earth System Modelling and otherpartners in the Global Organisation for Earth System SciencePortals (GO-ESSP). The authors wish to especially acknowledgeSamuel Somot (CNRM) for his fruitful discussions about regionalmodeling and the MedCORDEX database. Thanks are given toElisabeth Harader for some methodological ideas examined in theirPhD thesis. Special thanks are also given to Jean-Marc Lacave andGwenaëlle Hello from Météo-France for providing the daily and3 h pluviometer data, and to Céline Vargel for her useful work onthe comparison between SAFRAN and pluviometer data.

Edited by: Shraddhanand ShuklaReviewed by: two anonymous referees

References

Bechtold, P. and Bazile, E.: The 12–13 November 1999flash flood in southern France, Atmos. Res., 56, 171–189,https://doi.org/10.1016/S0169-8095(00)00097-1, 2001.

Boberg, F. and Christensen, J. H.: Overestimation of Mediterraneansummer temperature projections due to model deficiencies, Nat.Clim. Change, 2, 433, https://doi.org/10.1038/NCLIMATE1454,2012.

Boudevillain, B., Delrieu, G., Galabertier, B., Bonnifait, L., Bouil-loud, L., Kirstetter, P.-E., and Mosini, M.-L.: The Cévennes-Vivarais Mediterranean Hydrometeorological Observatorydatabase, Water Resour. Res., 47, doi:10.1029/2010WR010353,2011.

Bougeault, P.: A simple parameterization of the large-scale effects of cumulus convection, Mon. WeatherRev., 113, 2108–2121, https://doi.org/10.1175/1520-0493(1985)113<2108:ASPOTL>2.0.CO;2, 1985.

Cassou, C. and Cattiaux, J.: Disruption of the European cli-mate seasonal clock in a warming world, Nat. Clim. Change,https://doi.org/10.1038/NCLIMATE2969, 2016.

Cavicchia, L., Scoccimarro, E., Gualdi, S., Marson, P., Ahrens,B., Berthou, S., Conte, D., Dell’Aquila, A., Drobinski, P.,Djurdjevic, V., Dubois, C., Gallardo, C., Li, L., Oddo,P., Sanna, A., and Torma, C.: Mediterranean extreme pre-cipitation: a multi-model assessment, Clim. Dynam., 1–13,https://doi.org/10.1007/s00382-016-3245-x, 2016.

Christensen, J. H., Carter, T. R., Rummukainen, M., and Amana-tidis, G.: Evaluating the performance and utility of regional cli-mate models: the PRUDENCE project, Clim. Change, 81, 1–6,https://doi.org/10.1007/s10584-006-9211-6, 2007.

Clarke, L., Edmonds, J., Jacoby, H., Pitcher, H., Reilly, J., andRichels, R.: CCSP synthesis and assessment product 2.1, Part A:scenarios of greenhouse gas emissions and atmospheric concen-trations, US Government Printing Office, Washington DC, 2007.

Coustau, M.: Contribution à la prévision des crues sur le bassin duLez: modélisation de la relation pluie-débit en zone karstique etimpact de l’assimilation de débits, Université de Montpellier II,2011.

Dayon, G.: Evolution du cycle hydrologique continental en Franceau cours des prochaines décennies, Université de Toulouse, Uni-versité Toulouse III-Paul Sabatier, 2015.

Dee, D. P., Uppala, S. M., Simmons, A. J., Berrisford, P., Poli,P., Kobayashi, S., Andrae, U., Balmaseda, M. A., Balsamo, G.,Bauer, P., Bechtold, P., Beljaars, A. C. M., van de Berg, L., Bid-lot, J., Bormann, N., Delsol, C., Dragani, R., Fuentes, M., Geer,A. J., Haimberger, L., Healy, S. B., Hersbach, H., Hólm, E. V.,Isaksen, L., Kållberg, P., Köhler, M., Matricardi, M., McNally,A. P., Monge-Sanz, B. M., Morcrette, J.-J., Park, B.-K., Peubey,C., de Rosnay, P., Tavolato, C., Thépaut, J.-N., and Vitart, F.: TheERA-Interim reanalysis: configuration and performance of thedata assimilation system, Q. J. R. Meteorol. Soc., 137, 553–597,https://doi.org/10.1002/qj.828, 2011.

Déqué, M. and Somot, S.: Weighted frequency distributionsexpress modelling uncertainties in the ENSEMBLES re-gional climate experiments, Clim. Res., 44, 195–209,https://doi.org/10.3354/cr00866, 2010.

Déqué, M., Somot, S., Sanchez-Gomez, E., Goodess, C. M., Jacob,D., Lenderink, G., and Christensen, O. B.: The spread amongstENSEMBLES regional scenarios: regional climate models,driving general circulation models and interannual variabil-ity, Clim. Dynam., 38, 951–964, https://doi.org/10.1007/s00382-011-1053-x, 2012.

Diaconescu, E. P., Gachon, P., and Laprise, R.: On the remappingprocedure of daily precipitation statistics and indices used in re-gional climate model evaluation, J. Hydrometeorol., 16, 2301–2310, https://doi.org/10.1175/JHM-D-15-0025.1, 2015.

Drobinski, P., Da Silva, N., Panthou, G., Bastin, S., Muller, C.,Ahrens, B., Borga, M., Conte, D., Fosser, G., Giorgi, F., Güt-tler, I., Kotroni, V., Li, L., Morin, E., Önol, B., Quintana-Segui,P., Romera, R., and Torma, C. Z.: Scaling precipitation extremeswith temperature in the Mediterranean: past climate assessmentand projection in anthropogenic scenarios, Clim. Dynam., 1–21,https://doi.org/10.1007/s00382-016-3083-x, 2016.

Ducrocq, V., Aullo, G., and Santurette, P.: Les précipitations in-tenses et les inondations des 12 et 13 novembre 1999 sur le sudde la France, https://doi.org/10.4267/2042/36293, 2003.

Ducrocq, V., Nuissier, O., Ricard, D., Lebeaupin, C., and Thou-venin, T.: A numerical study of three catastrophic precipitat-ing events over southern France, II: Mesoscale triggering andstationarity factors, Q. J. Roy. Meteorol. Soc., 134, 131–145,https://doi.org/10.1002/qj.199, 2008.

Durand, Y., Brun, E., Mérindol, L., Guyomarc’h, G., Lesaf-fre, B., and Martin, E.: A meteorological estimation of rele-vant parameters for snow models, Ann. Glaciol., 18, 65–71,https://doi.org/10.3198/1993AoG18-1-65-71, 1993.

Hydrol. Earth Syst. Sci., 22, 673–687, 2018 www.hydrol-earth-syst-sci.net/22/673/2018/

A. Colmet-Daage et al.: Evaluation of uncertainties 685

Estupina, V.: Vers une modélisation hydrologique adaptée à laprévision opérationnelle des crues éclair: application à de petitsbassins versants du sud de la France, Toulouse, INPT, 2004.

Fowler, H. J., Ekström, M., Blenkinsop, S., and Smith, A.P.: Estimating change in extreme European precipitation us-ing a multimodel ensemble, J. Geophys. Res., 112, D18104,https://doi.org/10.1029/2007JD008619, 2007a.

Fowler, H. J., Blenkinsop, S., and Tebaldi, C.: Linking climatechange modelling to impacts studies: recent advances in down-scaling techniques for hydrological modelling, Int. J. Climatol.,27, 1547–1578, https://doi.org/10.1002/joc.1556, 2007b.

Frei, C., Schöll, R., Fukutome, S., Schmidli, J., and Vi-dale, P. L.: Future change of precipitation extremes inEurope: Intercomparison of scenarios from regional cli-mate models, J. Geophys. Res.-Atmos., (1984–2012), 111,https://doi.org/10.1029/2007JD008619, 2006.

Gao, X., Pal, J. S., and Giorgi, F.: Projected changes in mean andextreme precipitation over the Mediterranean region from a highresolution double nested RCM simulation, Geophys. Res. Lett.,33, https://doi.org/10.1029/2005GL024954, 2006.

Gaume, E., Livet, M., Desbordes, M., and Villeneuve, J.-P.:Hydrological analysis of the river Aude, France, flash floodon 12 and 13 November 1999, J. Hydrol., 286, 135–154,https://doi.org/10.1016/j.jhydrol.2003.09.015, 2004.

Gaume, E., Borga, M., Llasat, M. C., Maouche, S., Lang, M., andDiakakis, M.: Mediterranean extreme floods and flash floods.Into Hydro-meteorological extremes, chapter 3, The Mediter-ranean Region under Climate Change, A Scientific Update (co-ordinated by All Envi), 133–144, 2016.

Giorgi, F.: Climate change hot-spots, Geophys. Res. Lett., 33,https://doi.org/10.1029/2006GL025734, 2006.

Giorgi, F. and Lionello, P.: Climate change projections forthe Mediterranean region, Glob. Planet. Change, 63, 90–104,https://doi.org/10.1016/j.gloplacha.2007.09.005, 2008.

Giorgi, F., Jones, C., Asrar, G. R.: Addressing climate informationneeds at the regional level: the CORDEX framework, World Me-teorological Organization (WMO) Bulletin, 58, 175, 2009.

Giorgi, F., Torma, C., Coppola, E., Ban, N., Schär, C., and Somot,S.: Enhanced summer convective rainfall at Alpine high eleva-tions in response to climate warming, Nat. Geosci., 9, 584–589,https://doi.org/10.1038/ngeo2761, 2016.

Hagemann, S., Machenhauer, B., Jones, R., Christensen, O. B.,Déqué, M., Jacob, D., and Vidale, P. L.: Evaluation of water andenergy budgets in regional climate models applied over Europe,Clim. Dynam., 23, 547–567, https://doi.org/10.1007/s00382-004-0444-7, 2004.

Harader, E.: L’impact du changement climatique sur les événementshydrologiques extrêmes des petits bassins versants méditer-ranéens: le cas du bassin versant du Lez, Université de Toulouse,Université Toulouse III-Paul Sabatier, 2015.

Harris, I., Jones, P. D., Osborn, T. J., and Lister, D. H.: Up-dated high-resolution grids of monthly climatic observations– the CRU TS3. 10 Dataset, Int. J. Climatol., 34, 623–642,https://doi.org/10.1002/joc.3711, 2014.

Haylock, M. R., Hofstra, N., Klein Tank, A. M. G., Klok,E. J., Jones, P. D., and New, M.: A European daily high-resolution gridded data set of surface temperature and pre-cipitation for 1950–2006, J. Geophys. Res.-Atmos., 113,https://doi.org/10.1029/2008JD010201, 2008.

Herrera, S., Gutiérrez, J. M., Ancell, R., Pons, M. R.,Frías, M. D., and Fernández, J.: Development and analy-sis of a 50-year high-resolution daily gridded precipitationdataset over Spain (Spain02), Int. J. Climatol., 32, 74–85,https://doi.org/10.1002/joc.2256, 2012.

Herrera, S., Fernández, J., and Gutiérrez, J. M.: Update of theSpain02 gridded observational dataset for EURO-CORDEXevaluation: assessing the effect of the interpolation methodology,Int. J. Climatol., 36, 900–908, https://doi.org/10.1002/joc.4391,2016.

Jacob, D., Bärring, L., Christensen, O. B., Christensen, J. H., deCastro, M., Deque, M., Giorgi, F., Hagemann, S., Hirschi, M.,Jones, R., Kjellström, E., Lenderink, G., Rockel, B., Sánchez,E., Schär, C., Seneviratne, S. I., Somot, S., van Ulden, A., andvan den Hurk, B.: An inter-comparison of regional climate mod-els for Europe: model performance in present-day climate, Clim.Change, 81, 31–52, https://doi.org/10.1007/s10584-006-9213-4,2007.

Jacob, D., Petersen, J., Eggert, B., Alias, A., Christensen, O. B.,Bouwer, L. M., Braun, A., Colette, A., Déqué, M., Georgievski,G., Georgopoulou, E., Gobiet, A., Menut, L., Nikulin, G.,Haensler, A., Hempelmann, N., Jones, C., Keuler, K., Ko-vats, S., Kröner, N., Kotlarski, S., Kriegsmann, A., Martin,E., Meijgaard, E. van, Moseley, C., Pfeifer, S., Preuschmann,S., Radermacher, C., Radtke, K., Rechid, D., Rounsevell, M.,Samuelsson, P., Somot, S., Soussana, J.-F., Teichmann, C.,Valentini, R., Vautard, R., Weber, B., and Yiou, P.: EURO-CORDEX: new high-resolution climate change projections forEuropean impact research, Reg. Environ. Change, 14, 563–578,https://doi.org/10.1007/s10113-013-0499-2, 2013.

Katragkou, E., García-Díez, M., Vautard, R., Sobolowski, S., Za-nis, P., Alexandri, G., Cardoso, R. M., Colette, A., Fernandez,J., Gobiet, A., Goergen, K., Karacostas, T., Knist, S., Mayer,S., Soares, P. M. M., Pytharoulis, I., Tegoulias, I., Tsikerdekis,A., and Jacob, D.: Regional climate hindcast simulations withinEURO-CORDEX: evaluation of a WRF multi-physics ensemble,Geosci. Model Dev., 8, 603–618, https://doi.org/10.5194/gmd-8-603-2015, 2015.

Kjellström, E., Boberg, F., Castro, M., Christensen, J. H., Nikulin,G., and Sánchez, E.: Daily and monthly temperature and precip-itation statistics as performance indicators for regional climatemodels, Climate research (Open Access for articles 4 years oldand older), 44, 135, https://doi.org/10.3354/cr00932, 2010.

Knutti, R., Furrer, R., Tebaldi, C., Cermak, J., and Meehl,G. A.: Challenges in combining projections from multi-ple climate models, Journal of Climate, 23(10), 2739–2758,https://doi.org/10.1175/2009JCLI3361.1, 2010.

Kotlarski, S., Block, A., Böhm, U., Jacob, D., Keuler, K., Knoche,R., Rechid, D., and Walter, A.: Regional climate model simula-tions as input for hydrological applications: evaluation of uncer-tainties, Adv. Geosci., 5, 119–125, 2005.

Kotlarski, S., Keuler, K., Christensen, O. B., Colette, A., Déqué,M., Gobiet, A., Goergen, K., Jacob, D., Lüthi, D., van Meij-gaard, E., Nikulin, G., Schär, C., Teichmann, C., Vautard, R.,Warrach-Sagi, K., and Wulfmeyer, V.: Regional climate model-ing on European scales: a joint standard evaluation of the EURO-CORDEX RCM ensemble, Geosci. Model Dev., 7, 1297–1333,https://doi.org/10.5194/gmd-7-1297-2014, 2014.

www.hydrol-earth-syst-sci.net/22/673/2018/ Hydrol. Earth Syst. Sci., 22, 673–687, 2018

686 A. Colmet-Daage et al.: Evaluation of uncertainties

Kysel, J., Beguería, S., Beranová, R., Gaál, L., and López-Moreno, J. I.: Different patterns of climate change sce-narios for short-term and multi-day precipitation extremesin the Mediterranean, Glob. Planet. Change, 98, 63–72,https://doi.org/10.1016/j.gloplacha.2012.06.010, 2012.

Llasat, M. C., Llasat-Botija, M., and López, L.: A press databaseon natural risks and its application in the study of floods inNortheastern Spain, Nat. Hazards Earth Syst. Sci., 9, 2049–2061,https://doi.org/10.5194/nhess-9-2049-2009, 2009.

Llasat, M. C., Marcos, R., Llasat-Botija, M., Gilabert, J.,Turco, M., and Quintana-Seguí, P.: Flash flood evolution inNorth-Western Mediterranean, Atmos. Res., 149, 230–243,https://doi.org/10.1016/j.atmosres.2014.05.024, 2014.

Llasat, M. C., Marcos, R., Turco, M., Gilabert, J., and Llasat-Botija,M.: Trends in flash flood events versus convective precipitationin the Mediterranean region: The case of Catalonia, J. Hydrol.,541, 24–37, https://doi.org/10.1016/j.jhydrol.2016.05.040, 2016.

Lucas-Picher, P., Caya, D., de Elía, R., and Laprise, R.: Investiga-tion of regional climate models’ internal variability with a ten-member ensemble of 10-year simulations over a large domain,Clim. Dynam., 31, 927–940, https://doi.org/10.1007/s00382-008-0384-8, 2008.

Maraun, D., Osborn, T. J., and Rust, H. W.: The influence of synop-tic airflow on UK daily precipitation extremes, Part II, regionalclimate model and E-OBS data validation, Clim. Dynam., 39,287–301, https://doi.org/10.1007/s00382-011-1176-0, 2012.

Meinshausen, M., Smith, S. J., Calvin, K., Daniel, J. S., Kainuma,M. L. T., Lamarque, J. F., Matsumoto, K., Montzka, S. A., Raper,S. C. B., Riahi, K., Thomson, A., Velders, G. J. M., and vanVuuren D. P. P.: The RCP greenhouse gas concentrations andtheir extensions from 1765 to 2300, Clim. Change, 109, 213–241, https://doi.org/10.1007/s10584-011-0156-z, 2011.

Metzger, M. J., Bunce, R. G. H., Jongman, R. H. G., Mücher,C. A., and Watkins, J. W.: A climatic stratification of theenvironment of Europe, Glob. Ecol. Biogeogr., 14, 549–563,https://doi.org/10.1111/j.1466-822X.2005.00190.x, 2005.

Milano, M., Ruelland, D., Fernandez, S., Dezetter, A., Fabre, J., Ser-vat, E., Fritsch, J.-M., Ardoin-Bardin, S., and Thivet, G.: Currentstate of Mediterranean water resources and future trends underclimatic and anthropogenic changes, Hydrol. Sci. J., 58, 498–518, https://doi.org/10.1080/02626667.2013.774458, 2013.

Nuissier, O., Ducrocq, V., Ricard, D., Lebeaupin, C., and An-quetin, S.: A numerical study of three catastrophic precipitat-ing events over southern France, I: Numerical framework andsynoptic ingredients, Q. J. Roy. Meteorol. Soc., 134, 111–130,https://doi.org/10.1002/qj.200, 2008.

Nuissier, O., Joly, B., Joly, A., Ducrocq, V., and Arbogast, P.:A statistical downscaling to identify the large-scale circula-tion patterns associated with heavy precipitation events oversouthern France, Q. J. Roy. Meteorol. Soc., 137, 1812–1827,https://doi.org/10.1002/qj.866, 2011.

Mearns, L. O., Gutowski, W., Jones, R., Leung, R., Macgin-nis, S., Nunes, A., and Qian, Y.: A regional climatechange assessment program for North America, Eos, 90,https://doi.org/10.1029/2009EO360002, 2009.

Paxian, A., Hertig, E., Seubert, S., Vogt, G., Jacobeit, J., and Paeth,H.: Present-day and future mediterranean precipitation extremesassessed by different statistical approaches, Clim. Dynam., 44,845–860, https://doi.org/10.1007/s00382-014-2428-6, 2015.

Prein, A. F., Gobiet, A., Truhetz, H., Keuler, K., Goergen, K., Teich-mann, C., Maule, C. F., Van Meijgaard, E., Déqué, M., Nikulin,G., and others: Precipitation in the EURO-CORDEX 0.11◦ and0.44◦ simulations: high resolution, high benefits, Clim. Dynam.,46, 383–412, https://doi.org/10.1007/s00382-015-2589-y, 2016.

Quintana Seguí, P., Ribes, A., Martin, E., Habets, F., andBoé, J.: Comparison of three downscaling methods insimulating the impact of climate change on the hydrol-ogy of Mediterranean basins, J. Hydrol., 383, 111–124,https://doi.org/10.1016/j.jhydrol.2009.09.050, 2010.

Quintana-Seguí, P., Le Moigne, P., Durand, Y., Martin, E., Habets,F., Baillon, M., Canellas, C., Franchisteguy, L., and Morel, S.:Analysis of Near-Surface Atmospheric Variables: Validation ofthe SAFRAN Analysis over France, J. Appl. Meteor. Climatol.,47, 92–107, https://doi.org/10.1175/2007JAMC1636.1, 2008.

Quintana-Seguí, P., Habets, F., and Martin, E.: Comparison of pastand future Mediterranean high and low extremes of precipita-tion and river flow projected using different statistical down-scaling methods, Nat. Hazards Earth Syst. Sci., 11, 1411–1432,https://doi.org/10.5194/nhess-11-1411-2011, 2011.

Quintana-Seguí, P., Peral, C., Turco, M., Llasat, M. C., andMartin, E.: Meteorological Analysis Systems in North-EastSpain: Validation of SAFRAN and SPAN, J. Environ. Inf., 27,https://doi.org/10.3808/jei.201600335, 2016a.

Quintana-Seguí, P., Turco, M., Herrera, S., and Miguez-Macho, G.:Validation of a new SAFRAN-based gridded precipitation prod-uct for Spain and comparisons to Spain02 and ERA-Interim,https://doi.org/10.5194/hess-2016-349, 21, 2187, 2017.

Ramis, C., Llasat, M. C., Genovés, A., and Jansà, A.:The October-1987 floods in Catalonia: Synoptic andmesoscale mechanisms, Meteorol. Appl., 1, 337–350,https://doi.org/10.1002/met.5060010404, 1994.

Reifen, C. and Toumi, R.: Climate projections: Past perfor-mance no guarantee of future skill?, Geophys. Res. Lett., 36,https://doi.org/10.1029/2009GL038082, 2009.

Raynaud, F., Borrell-Estupina, V., Pistre, S., Van-Exter, S., Bour-geois, N., Dezetter, A., and Servat, E.: Combining hydraulicmodel, hydrogeomorphological observations and chemical anal-yses of surface waters to improve knowledge on karst flash floodsgenesis, Proc. Int. Assoc. Hydrol. Sci., 369, 55–60, 2015.

Riahi, K., Grübler, A., and Nakicenovic, N.: Scenarios of long-term socio-economic and environmental development under cli-mate stabilization, Technol. Forecast. Soc. Change, 74, 887–935,https://doi.org/10.1016/j.techfore.2006.05.026, 2007.

Ricard, D., Ducrocq, V., and Auger, L.: A climatology of themesoscale environment associated with heavily precipitatingevents over a northwestern Mediterranean area, J. Appl. Mete-orol. Climatol., 51, 468–488, https://doi.org/10.1175/JAMC-D-11-017.1, 2012.

Schär, C., Ban, N., Fischer, E. M., Rajczak, J., Schmidli, J., Frei, C.,Giorgi, F., Karl, T. R., Kendon, E. J., Tank, A. M. K., O’Gorman,P. A., Sillmannn, J., Zhang, X., and Zwiers, F. W.: Percentileindices for assessing changes in heavy precipitation events,Clim. Change, 137, 201–216, https://doi.org/10.1007/s10584-016-1669-2, 2016.

Themeßl, M. J., Gobiet, A., and Heinrich, G.: Empirical-statisticaldownscaling and error correction of regional climate models andits impact on the climate change signal, Clim. Change, 112, 449–468, https://doi.org/10.1002/joc.2168, 2012.

Hydrol. Earth Syst. Sci., 22, 673–687, 2018 www.hydrol-earth-syst-sci.net/22/673/2018/

A. Colmet-Daage et al.: Evaluation of uncertainties 687

Tramblay, Y., Ruelland, D., Somot, S., Bouaicha, R., and Servat,E.: High-resolution Med-CORDEX regional climate model sim-ulations for hydrological impact studies: a first evaluation ofthe ALADIN-Climate model in Morocco, Hydrol. Earth Syst.Sci., 17, 3721–3739, https://doi.org/10.5194/hess-17-3721-2013,2013.

Trapero, L., Bech, J., Duffourg, F., Esteban, P., and Lorente,J.: Mesoscale numerical analysis of the historical Novem-ber 1982 heavy precipitation event over Andorra (EasternPyrenees), Nat. Hazards Earth Syst. Sci., 13, 2969–2990,https://doi.org/10.5194/nhess-13-2969-2013, 2013.

van der Linden P. and J. F. B. Mitchell (eds.): ENSEMBLES: Cli-mate Change and its Impacts: Summary of research and re-sults from the ENSEMBLES project. Met Office Hadley Centre,FitzRoy Road, Exeter EX1 3PB, Uk, 160 pp, 2009.

Van Vuuren, D. P., Edmonds, J., Kainuma, M., Riahi, K., Thom-son, A., Hibbard, K., Hurtt, G. C., Kram, T., Krey, V.,Lamarque, J.-F., Masui, T., Meinshausen, M., Nakicenovic,N., Smith, S. J., and Rose, S. K.: The representative con-centration pathways: an overview, Clim. Change, 109, 5–31,https://doi.org/10.1007/s10584-011-0148-z, 2011.

Vidal, J.-P., Martin, E., Franchistéguy, L., Baillon, M., and Soubey-roux, J.-M.: A 50-year high-resolution atmospheric reanalysisover France with the Safran system, Int. J. Climatol., 30, 1627–1644, https://doi.org/10.1002/joc.2003, 2010.

www.hydrol-earth-syst-sci.net/22/673/2018/ Hydrol. Earth Syst. Sci., 22, 673–687, 2018